Apple v. OpenAI: The AI Agent Caught in a Trade-Secrets Fight

Apple has escalated its trade-secrets suit against OpenAI, filing a motion for expedited discovery and accusing the company of actively destroying evidence while the case is pending. The filing, reported by Bloomberg and covered by The Verge, centers on former Apple engineer Chang Liu, who now works at OpenAI.

What Apple alleges

According to the court filing, Liu downloaded a confidential Apple circuit schematic in March 2026 — about two months after leaving Apple for OpenAI. Apple says forensic analysis of a company-issued MacBook Liu kept after departing shows he used that schematic in his OpenAI work, running it through LTspice, a circuit-simulation tool, and also used an OpenAI internal tool that shares its name with an Apple engineering application.

The detail that stands out for builders: Liu allegedly didn't run the simulations by hand. Per the filing, he told colleagues that 'in the past hour, his AI agent learned how to run LTspice, look at result, tune compensation parameter' — describing an agent he'd set up to iterate on the disputed schematic autonomously.

The evidence-destruction claim

Apple's motion says that once Liu learned Apple was investigating him, he enlisted an OpenAI colleague, Yu-Ting Peng, to help destroy evidence — and that she confirmed she would comply. Apple argues this is why it needs expedited discovery now, rather than waiting for the normal litigation timeline: if the underlying material disappears, a later finding of liability won't undo the damage. OpenAI disputes Apple's characterization of events, including Apple's claim that Liu exploited an authentication bug to retain file access after leaving.

Apple is asking the court for two things: expedited fact-finding, and a preliminary injunction blocking OpenAI from developing hardware based on the disputed technology while the case proceeds. Both sides argue the motion before a federal judge on October 1.

Why this matters beyond one lawsuit

Talent has been flowing between Apple and OpenAI in both directions for years, and this case is a reminder that hardware and silicon expertise — not just model weights — is now squarely inside AI companies' IP battles. But the more immediate lesson for engineering teams shipping with AI agents in the loop is procedural: an agent's action history is turning into the kind of artifact that shows up in a discovery filing, alongside chat logs and file-access timestamps.

If an autonomous agent touched a file, ran a tool, or iterated on a design, that trail is potentially as material as a code diff — and, per this filing, potentially something a defendant would want gone. Teams building agentic workflows against sensitive IP should treat agent logs with the same retention and audit discipline as source control, not as disposable scratch output.

What happens next

Nothing is decided yet — this is Apple's version of events in a motion, not a court finding, and OpenAI has pushed back on parts of it. The October 1 hearing will determine whether Apple gets the accelerated discovery timeline and injunction it's asking for.